Improving YOLOv8 Deep leaning model in rice disease detection by using Wise - IoU loss function

Abstract

This paper presents an improved method for a deep learning model applied to the detection of diseases in rice crops. Early detection and prevention of pests and diseases are essential to ensure effective crop productivity. The YOLOv8 deep learning model was employed to detect three common diseases in rice leaves: leaf folder, rice blast, and brown spot. To enhance the model's performance, we replaced the default CIoU loss function in YOLOv8 with WIoU, achieving an overall accuracy of 89.2%, with an improvement of 4.5% on mAP@50 and 4.4% on mAP@50-95. These results demonstrate promising potential for improving the performance and reliability of deep learning models in agricultural applications.